Probabilistic predictive warning system for process operations

该项目旨在开发一种概率预测方法来设计和管理过程工业中使用的早期预警系统。该方法的特点和新颖之处在于在预警设计的各个阶段（即分配、预警、排序和管理）中使用概率模型进行预警通报，并将风险纳入预警设计的各个阶段。此外，预报将用于发布警告。***行业显著受益于使用预测进行控制。另一方面，相当多的资源用于风险评估。然而，目前的报警系统不是基于风险的；预测的使用也受到限制。需要进行研究工作，使工厂报警系统成为防止事故的有效工具。***我们计划开发一个基于量化风险和从工厂数据预测事件的综合预警系统。目前的工厂报警系统使用确定性设置，其中异常描述没有任何参考其可能性。我们的目标是开发工具，以可能性来描述不想要的事件，并通过考虑解释观察结果的一组可能的故障组合来进行诊断。将开发方法来量化工厂模型的不确定性，这些模型是由模型结构的假设和过程数据中的噪声产生的。还将讨论与风险模型有关的不确定性。事件之间的相关性和依赖性以及它们的后果将被量化，并在预警系统的设计中加以考虑。我们还想提出一种警报设计理念，以补充纳入风险的现有标准。将开发一个可与现有工厂警报系统集成的原型警报系统。新系统将在试点工厂和工业环境中进行测试。加工工业急需一种改进的报警系统。操作员经常被警报淹没，无法履行他们的主要职责。风险是一种被充分理解和高度重视的度量。因此，开发的方法有望在过程工业中得到广泛的应用。***通过开发技术和工具，确保工业工厂顺利，无事故运行，该计划将直接促进工业部门的生产力。因此，它将有助于节省数十亿美元的损失造成的异常情况的行业。随着工具和方法的发展，一些高素质的人才将在这个项目中接受培训。HQP将在开发工具及其在实际工厂中的实施方面接受培训；因此，HQP将成为有关行业的宝贵资产。最后，所开发的方法将有助于在过程设施之外的应用中进行风险评估，例如用于健康监测、经济活动和自然灾害。***********

This program aims to develop a probabilistic predictive approach to the design and management of early warning systems used in process industries. The special features and novelty of the proposed approach is in the use of probabilistic models for alarm annunciation and the use of risk in every stage of warning design, namely, allocation, annunciation, prioritization and management. In addition, predictions will be used for issuing warnings. ***Industries significantly benefit from using predictions for control. On the other hand, considerable resources are used for risk assessment. However, current alarm systems are not risk-based; use of predictions is also restricted. Research efforts are required to make plant alarm systems an effective tool in preventing accidents. ***We plan to develop a comprehensive early warning system based on quantified risk and prediction of events from plant data. The current plant alarm systems use a deterministic setting where abnormality is described without any reference to its likelihood. Our objective is to develop tools to describe unwanted events in terms of likelihood and perform diagnosis by considering a set of possible combinations of failures that explain the observations. Methodologies will be developed to quantify uncertainties with plant models which arise from assumptions on model structures and noises in process data. Uncertainties associated with risk models will also be developed. Correlations and dependencies among events as well as their consequences will be quantified and taken into account in the design of warning system. We also want to propose an alarm design philosophy to complement the existing standards in incorporating risk. A prototype alarm system will be developed that will be integrate-able with the existing plant alarm systems. The new system will be tested in pilot plants and industrial settings.***Process industries are in dire need of an improved alarm system. Operators are often overwhelmed by alarms and are unable to perform their main responsibilities. Risk is a measure that is well understood and highly valued. Hence, the developed methodology is expected to find widespread utility in process industries.***By developing technologies and tools that ensure smooth, accident-free operation of industrial plants, this program will contribute directly to the productivity of the industrial sector. Thus it will help to save billions of dollars in losses caused by abnormal situations in industries.***Along with the development of tools and methods, a number of highly qualified personnel will be trained in this program. The HQP will be trained on developing the tools as well as its implementation to real plants; thus the HQP will be valuable assets for the concerned industry. Finally, the developed methods will be useful for risk assessment in general with applications outside process facilities, e.g. for health monitoring, economic activities and natural disasters.***********